4.new-mj-hawaii

Download Report

Transcript 4.new-mj-hawaii

Joint Techs / APAN
Honolulu
Mark Johnson
MCNC (NCREN, NCNI, NCLR, …)
[email protected]
How can an R&E network afford to
build an advanced network?
 Use the obvious strategy of obtaining
donations from providers and equipment
vendors and the use of grants
 Make more efficient use of the available
(scarce) resources - MORPHnet
Infrastructure
Experimental
Production L3
or breakable
gear
Production IP service (Cisco
Experimental
L3 gear
Production L2COTS routers) - 10GE and
or breakable
3 gear
Production Ethernet service
Experimental
1GE ports
L2-3 gear
(Cisco COTS switches) - 1GE
or breakable
ports
L1-3 gear
Production point-to-point wave service (Cisco COTS DWDM gear) - 10GE, 1GE, OC192
and OC48 waves
Production fiber (1st pair)
NLR operated
NLR or its production customer or researcher operated
Research use
Production use
Production
fiber (2nd pair)
Production and experimental infrastructure(MORPHnet concept) and their use
Research needing its own dark fiber full
spectrum and/or deployment of
breakable L1 gear, e.g., optical packet
switching, IP-optics unified control
plane, 100GE optics
Production use of dedicated (multiple)
10G bandwidth, e.g., DTF/ETF cluster
supercomputers "backplane"
interconnect, federal agency mission
use, international connections transit
Research needing its own L1 links
and/or dedicated 10G bandwidth, e.g.,
very large MTU performance, XTP
implementation
Production use for cases where shared
IP service is not acceptable but also
dedicated 10G waves not needed
either, e.g., remote instrument control
Research needing its own L2 links with
the capability to do complex topologies
but where speed is not the primary
focus and 1GE or lower ports are
sufficient, e.g., multicast routing
Production use for higher ed and K-12
AUP-free commodity Internet access
and inter-GigaPoP transit backup
Research based on measurements of
real user Internet traffic (and not just
univ-to-univ traffic) and visibility into
Internet BGP for the first time since
NSFnet
Infrastructure Use Examples
What does a user want from
an optical network?
 An end-to-end path (lightpath) where the
endpoints are not defined by the limits of a
single carrier’s network
Lightpaths
 a lightpath is defined to be a
fixed bandwidth connection
between two network elements,
such as IP routers or ATM
switches, established via the
optical network
 Ietf draft on lightpath attributes
Lightpath attributes
 It is assumed that a lightpath will have a
number of attributes that describe it such
as framing, bandwidth, etc
 Canarie asserts that across a given AS a
lightpath may be abstracted to look like a
single (possibly blocking) cross-connect
switch interface.
working examples of
Lightpaths
 All Optical wavelength on WDM system
 SONET channel
 Point to point ethernet
 ATM CBR circuit
 MPLS LSR with defined QoS
 Fiberchannel
 SMPTE 259
 G.709 (Digital Wrapper)
Problems
 Intra-domain
 Provisioning of network capacity across
network elements within an AS
 O&M
 Inter-domain
 Provisioning of network capacity across
multiple AS’s
 O&M
 In this environment the user has to handle
performance and fault management
Lightpath
Carrier A
Carrier B
Carrier C
User desires red path but must negotiate and manage provisioning of green, orange, and blue paths
Approaches
 Methods of defining, provisioning, and
modifying existing services within a
management domain
 G.ASTN
 GMPLS
 Methods of linking paths from multiple domains
 UCLP
 Non-traditional techniques for provisioning
capacity between endpoints
 OBS/JIT
GMPLS
 Generalized MPLS signaling to identify the
following path types:
 Traditionally statistically multiplexed labeled paths
such as ATM or Ethernet
 Time division multiplexed paths such as SONET
where timeslots are the label
 Frequency division multiplexed services such as
wavelengths where frequency is the label
 Space division multiplexed services such as fibers in
a bundle where position in the real world is the label
Division of labor
 Control plane
 Signaling, routing, Protection /restoration
 Transport
 Adaptation, Aggregation, Discovery,data
integrity, transmission
 Management
 Management of Faults,
configuration/provisioning, accounting,
performance measurement, security
Division of labor
Network Topology Map
Control plane
based on IP Routing
Forwarding plane
IP
ATM
Today
Drawing poorly copied from Cisco Systems
Future
Optical
GMPLS Protocol Diagram
LMP
RSVP-TE
CR-LDP-TE
UDP
TCP
Adaptation Layer
SONET
Wavelength
Switching
FIBER
BGP
OSPF-TE
IP
MAC/GE
ATM
Frame
Relay
UCLP
 Canarie is developing a system including
protocols and directories and registration
mechanisms Addresses interdomain
issues:
 Registration of available path components
 Directory service for those components
 Provisioning of end to end path which
could use intra-domain tools such as
GMPLS
JIT/OBS view of Optical Network Dilemma
 Goal: Lower cost by:
 Minimizing OEO
 Creating larger transparency islands
 But:
 Dedicated  is overkill (expensive)
 Low speed apps. need fine grain mux capability
 And:
 Existing fine grain multiplexing today requires
electronics hence OEO conversion
Technology gap
Gauger et al., “Determining offset
times in optical burst switching networks”,
COST 266, Zagreb, June 2001.
 Requires optical buffers
 Immature, expensive, low density
 Buffers in net lead to complexity
 IP is a COMPLEX protocol
 Hardware implementations only recently
 Creates cost and technology barrier
Optical Cell Switching
 TDM dWDM
 All wavelengths on fiber switched together
 Pluses and minuses




Simpler core network
Need chromatic time correction
Requires frame synchronization
Low utilization of wavelenghts
 Lucent is major proponent
Three Competing Ideas
Need for
optical
buffering
Synchronization
Relative
timeline to
commercial
viability
Relative
complexity
comments
OPS critical
No
Not in our
career lifetime
Complex
switching and
protocol
Requires
optical logic
OCS no
yes
5 to 10 years
Simplest
switch core,
with most
complex line
cards
Requires
chromatic
and frame
allignment,
low 
utilization
OBS No longer
seen as
necessary
or desirable
Not
required
for JIT,
limited
sync for
JET
2 years with
20 ms, 7 years
with 10 ns
reconfiguration
Simple line
with cards
modestly
complex
switch core
Requires 
demultiplexin
g and
conversion
JIT Fundamental Values
 Low latency is first priority
 Tell and go vs. tell and wait
 May sacrifice link utilization
JET and Horizon
 Aggressive protocol simplification
 A pox on buffers (optical delay lines)
Leads to un-necessary protocol and switch complexities
 Leads to greater link speed and lower latency
 Keep data in optics
 No legacy assumptions
 Result: high throughput, min. latency and jitter
JIT - OBS Approach
 Switched light path network
 Large all-optical island
 No buffers in data channel
 Avoids immature device technology
 No buffer overflow in network
 Data and signaling channel isolation
 Single out of band signaling channel per fiber
 Signaling msgs. undergo OEO, processed by
intermediate nodes
 Network intelligence is concentrated at edge
 SIMPLE protocol implemented in hardware
ECOnet
 Create a confederation of fiber linked NRT projects:
 BOSnet:
 MIT Lincoln Labs
 dark fiber Boston to Washington DC
 ATDNet:
 Naval Research Lab (& others)
 dark fiber within the Washington D.C. metro area
 ECO-South (proposed):
 MAX/MCNC/SOX
 Dark fiber from Washington to Research Triangle Park and then
to Atlanta
East Coast Optical Network
ECOnet
MITLL
Boston, MA
BOSnet
MAX/ATDNet
Washington, DC
ECO-South
MCNC/NCREN
Raleigh, NC
GaTech/SOX
Atlanta, GA
Illustrates need to evaluate Entire system
 Fiber, amps, DCUs, maintenance and
Rent can become dominant costs
Two fiber routes
Route A
Washington, D.C. to Richmond
Richmond to Raleigh
Raleigh to Charlotte
Charlotte to Atlanta
TOTAL
Route B
Miles
129
170
208
260
ILA's
-
1
2
3
3
767
Miles
Washington, D.C. to Raleigh
476
Raleigh to Atlanta
563
TOTAL
3R's
1,039
9
3R's
-
ILA's
9
12
21
Fiber cost
Route A
Washington, D.C. to Richmond
Richmond to Raleigh
Raleigh to Charlotte
Charlotte to Atlanta
Number of Fiber mile
Fibers
Price
2
2
2
2
$
$
$
$
500
500
500
500
TOTAL
20 YR IRU
Price
$
$
$
$
fiber totals
129,000
170,000
208,000
260,000
$
299,000
$
468,000
$ 767,000
$
767,000
Route B
Washington, D.C. to Raleigh
2 $
-
$
-
$
-
Raleigh to Atlanta
2 $
-
$
-
$
-
-
TOTAL
-
-
Amps, colo, maintenance
Route A
Washington, D.C. to Richmond
Richmond to Raleigh
Raleigh to Charlotte
Charlotte to Atlanta
ILA Racks
(20ADC)
ila rent
POP
Racks
(20ADC)
maint
POP rent
amp HW
amps
$
600
$ 21,600
800
9600 $
29,900
$ 126,000
$
504,000
$
600
$ 43,200
800
9600 $
46,800
$ 126,000
$
882,000
$
76,700
TOTAL
$ 64,800
$ 1,386,000
Route B
Washington, D.C. to Raleigh
$
400
$ 43,200
Raleigh to Atlanta
$
400
$ 57,600
100,800
TOTAL
$
800
9600
0 $ 126,000
$ 1,260,000
800
9600
0 $ 126,000
$ 1,638,000
2,898,000
5 year total
Route A
Washington, D.C. to Richmond
Richmond to Raleigh
Raleigh to Charlotte
Charlotte to Atlanta
TOTAL
Route B
Washington, D.C. to Raleigh
total year 1 total year 2 total year total year 4 total year 5
(dim)
(dim)
3 (dim)
(dim)
(dim)
total yr 1-5
624,900
$ 120,900
$ 120,900
$ 120,900
$ 120,900
1,108,500
1,075,200
$ 183,600
$ 183,600
$ 183,600
$ 183,600
1,809,600
1,700,100
304,500
304,500
304,500
304,500
2,918,100
1,312,800
$
43,200
$ 43,200
$
43,200
$
43,200
1,485,600
1,705,200
3,018,000
$
57,600
100,800
$ 57,600
100,800
$
57,600
100,800
$
57,600
100,800
1,935,600
3,421,200
Raleigh to Atlanta
TOTAL
NCNI WDM Network
Duke
DCU
15454 Node
UNC
DCU
Network
Management
Access
DCU
DCU
Cisco RTP
MCNC
MCNC
Dispersion
Compensation Unit
DCU
DCU
Engineering Notes:
•Ring Circumference = 178.1099km
•SMF-28 Fiber
NCSU
DCU
RLGH
EDFA
amplifier
SMJ 5-27-03